Fix wrong AI

Three simple tips to fix wrong AI

Here are three easy tips to detect and fix when AI is wrong.  AI can 'hallucinate'. You should always be double checking the output before using it for something important.  

Number one is cross referencing across multiple AI-models. Say, you use ChatGpt for your original prompt, go ahead and ask Gemini, go ahead and ask DeepSeek, and see if the answers converge. 

Number two is a pattern called reflection. Ask AI to reflect on its previous answer and how it could be improved, ask it to cite its sources, and ask to fact check its previous answer. 

Number three, explicitly ask AI: if you are less than 90% confident about an answer or sentence 'flag' it with "UNCERTAIN" in caps. 

That's still not perfect, it still may miss certain things. So, when accuracy is important do all three of these tips. 

LLM = Large Language Model = AI-model/system